WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

Analysis of Attribute Acceptance Sampling Properties

SEJID TENJAK
Faculty of Electrical Engineering and Computing
University of Zagreb
Unska 3, 10000 Zagreb
CROATIA
[email protected]

IVAN BANOVAC
School of Medicine
University of Zagreb
alata 3, 10000 Zagreb
CROATIA
[email protected]

Abstract: - Attribute sampling is a powerful quality inspection tool. The Lot Acceptance Sampling Plans
(LASPs) for sampling by attributes represent a valuable method of quality inspection. Significant LASPs such
as the c=0, Dodge-Romig, Philips as well as ANSI/ASQC Z1.4 which replaced MIL-STD 105E have
comprehensive procedures and statistical base. A single LASP can be implemented efficiently if inspection of
incoming lots is unpractical and expensive due to the destructive and/or time-consuming procedure. The
implementation of LASPs is useful when a large number of purchased items is inspected daily in any complex
industrial branch, such as electric-power or gas, but even in large medical laboratories. The significant issues of
a complex attribute acceptance sampling statistical frame are considered in the paper. It is demonstrated how
the lot size differs over different double sampling plans.
Key-Words: - Attribute acceptance sampling, Attribute case, Quality inspection
especially useful if the laboratory supply chain (as a
complex part of the national health system) or large
laboratories procured a large number of items for
which the top characteristics of quality are often
demanded. Furthermore, the tenderers that receive
the contract award decisions in the public
procurement procedures become new suppliers of
goods. If a supplier is new, the quality of goods is
always questionable from the purchaser's point of
view. When a new supplier provides a shipment of a
large number of items, the following question will
appear: is the shipment good enough to put into
stock? A purchaser must decide whether to accept or
reject the received shipment.
Achieving quality in a medical laboratory
requires the use of different tools, such as
calibrations, procedure manuals, maintenance
schedules and quality inspection which encompass
the operational techniques and activities used to
fulfill prescribed requirements for quality. Testing
reliability can be achieved using quality control
materials as well as the product quality control.
Although the set for evaluating analytical run
quality for medical laboratories, i.e. six basic rules
created by J.O. Westgard [1]-[2], is often used in the
laboratory quality control, it is important to
implement a valid sampling acceptance procedure
when a purchaser evaluates large lots of materials or
products with the aim of fulfilling all requirements
for quality.
The acceptance sampling implementation by
using the most important LASPs for attribute case,

1 Introduction
Developed techniques are used for sampling as well
as quality control and functionality diagnostics of
the primary and secondary equipment in complex
industries, such as electric-power or gas. In common
inspection, it is necessary to control a large number
of items before installation. It is necessary to use the
high reliable acceptance sampling method because
the assembled items have to be of top-quality level
to assure the safety of the electric-power or gas
system. At first sight, it seems that 100% inspection
would be an appropriate approach to ensure high
reliable items. Obviously, 100% inspection is
unpractical and expensive if the inspection method
of a large number of items is destructive and/or
time-consuming. Therefore, an alternative approach
should be implemented. In such a case the
acceptance sampling is employed and a chosen Lot
Acceptance Sampling Plan (LASP) represents the
best option because it enables undoubted conclusion
on the lot's disposition based on counting the
number of defectives in a sample picked from a lot
randomly.
The implementation of acceptance sampling can
be also useful in other sectors. The authors advocate
for using LASPs in medicine, especially in the
medical laboratory procurement. Generally, an
efficient procurement of laboratory items impacts
the quality of laboratory services. Hence, the
procurement of laboratory items (such as
consumables, glass wares, instruments) is an
important issue. The use of LASPs could be

E-ISSN: 2224-2678

720

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

The rest of the paper is structured as follows.

Section 2 provides the categories of LASPs. Section
3 describes the average sample number. Section 4
summarizes the properties of the operating
characteristic curve. The average amount of
inspection per lot is described in section 5. The
parameters' comparison for different LASPs is
presented in section 6. Finally, section 7 draws some
relevant conclusions.

such as Dodge-Romig [3]-[5], Philips [6], MIL-STD

105E [7] (replaced by: ANSI/ASQC Z1.4; ISO
2859) and the c=0, is strongly statistical based,
therefore this use is reliable.
Sampling plans developed for sampling by
variables, such as the BENDIX and LOT-PLOT, are
not considered in this paper which deals with
sampling by attributes. Inspection by attributes is a
kind of quality inspection in which any a unit is
classified either as conforming or nonconforming,
with respect to a defined set of requirements.
A process in which a sample (ni) is picked at
random from the lot is carried out by a LASP.
Random sampling ensures that all samples of size
(n) are equal. Each LASP defines a number of units
that should be sampled from a lot and how many
defects are allowed in the chosen sample. LASP
represents a sampling scheme and a set of rules for
making decisions whether a lot should be accepted
or rejected [8]. A decision can be made regarding
the disposition of the lot (N) on the basis of the
resulting information. A point to emphasize is that
the aim of acceptance sampling is to decide whether
or not the lot is likely to be acceptable, not to
estimate the lot quality.
An explanation of the difference between
sampling by attributes and sampling by variables is
necessary. The characteristic of sampling by
attributes is that the item inspection leads to a binary
result: either the item is conforming or
nonconforming. On the other hand, sampling by
variables is characterized by the item inspection
leading to a continuous measurement. Sampling by
attributes is more common than sampling by
variables regardless of the industry sector in which
it is used. For instance, many components come in
large lots in the case of the electricpower industry,
which is often a subject of regulation [9]-[11].
Hence, it is not possible to provide 100% inspection.
In such an industry branch, all components have to
be of top-quality to ensure the reliability of the
power system. In this sense, it is necessary to select
a suitable number of components that have to be
inspected before installation regardless of using the
components for everyday operation or overhaul.
Some important references that deal with
acceptance sampling are [12]-[14].
Quality inspection used in industrial branches
can also include other approaches. However, the
combination of both acceptance sampling and
process control can often ensure an efficient quality
inspection.
A contribution of this research is in expressing
how the right chosen LASP ensures reliability of
acceptance of a large number of quality items.

E-ISSN: 2224-2678

2 The categories of LASPs

LASPs consist of single, double and multiple
sampling plans. Making a choice between single and
multiple sampling plans is a matter of deciding
whether the average sampling savings gained by the
multiple plans justifies an additional complexity of
these plans and the uncertainty of not knowing how
much inspection will be done on a daily basis.
Single sampling plans
A single sampling plan is denoted as (n,c). It is the
most common plan used in quality control in which
one sample of items (n) is selected at random from a
lot (N). The disposition of the lot depends on the
resulting information: the lot is rejected if there are
more than c defectives (note: c is the acceptance
number). This plan has a significant disadvantage,
i.e. it requires a large number of samples. Hence,
some alternative double or multiple plans should be
considered in the case of destructive testing of
expensive items when the number of samples is the
most important criterion.
Double sampling plans
A double sampling plan has good efficiency [15].
Another sample is taken in the double sampling
procedure if the first sample taken from the lot is not
informative enough, i.e. the sample (n2) is taken if
the result of the sample (n1) is not conclusive with
regard to accepting or rejecting. The lot is not
acceptable if the number of defectives is under the
acceptance number (c2). If the number of defectives
is between the acceptance numbers (c1) and (c2), the
sample (n2) is taken in order to establish whether the
total number of defectives in both samples is bigger
or smaller than the acceptance number (c2). The
final decision is based on this information.
One Philips plan can be a single (used if lot size
is up to 1,000) or multiple (for example, a double
sampling plan is used if the lot size is over 1,000).
Nevertheless, Philips is typically a highly
characteristic double sampling plan for which:
there are sample sizes n1 and n2 (with the
precondition: n2 = 2n1) and a point of neutral
quality pn,

721

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

the acceptance number (c) is the maximum

number of defectives in sample (n) in the case of
acceptance of lot (N),
the relation n1, c1, c2 = f(N) is valid.
the values of the Acceptable Quality Level (pa),
pt and the Average Outgoing Quality Level
(AOQL) are necessary for drawing the operating
characteristic curve (their values are defined by
the Philips tables: for example, if a lot size is
1,100 and the acceptable quality level (pa) is
1.0%, the corresponding pt is 5.8% with the
AOQL of 1.6%).
The usefulness of Philips plan can be
demonstrated by using an example. Let some lots be
delivered to control. The quality control department
will inspect the shipments by using an appropriate
sampling plan. According to the contract, the
supplier is obligated to send lots of 1,100 units with
the Acceptable quality level of 1.0% and Philips
double sampling plan is used. To carry out the
inspection procedure more easily, it is useful to
draw a sampling scheme for this case (Fig. 1). The
sampling scheme of double Philips sampling plan is
drawn using the values taken from the table of this
plan, for N = 1,0012,000 and pa = 1%, as follows:
AOQL = 1,6%,
pn = 3% and pt = 5,8%.
Furthermore, by using pn = 3% the following values
are taken from another table of Philips sampling
plan: n1 = 45, c1 = 0 and c2 = 3.

Multiple sampling plans

A multiple sampling plan introduces an extension of
the double sampling plan because additional
samples can be drawn after the second sample in the
multiple sampling procedure. Consequently, more
than two samples are needed to reach a conclusion.
In fact, multiple sampling plans were introduced to
give a questionable lot another chance to be
accepted.
The samples are taken at random and the
sampling procedure is reliable. Although multiple
sampling plans are complex and the possibility of
error is greater, they can reduce the number of
samples. These plans have smaller sample sizes and
a shorter examination period, which is very useful in
the case of complex products [16].
Multiple sampling plans are usually presented in
a tabular form. The multiple sampling procedures
commence with taking a random sample of size (n1)
from a large lot of size (N) and counting the number
of defectives (d1). If d1 a1, the lot is accepted.
Furthermore, if d1 r1, the lot is rejected. Another
sample should be taken if a1 < d1 < r1. The firstsample-procedure is repeated sample by sample if
subsequent samples (n2) are required. The total
number of defectives (Nr) found at any stage of
multiple sampling can be calculated as the sum of
all numbers of defectives at each stage (dj), where j
ranges from 1 to the total number of stages in
multiple sampling (k).
It is a rule for each stage of multiple sampling
procedure that the number of defectives is compared
to the acceptance number (ai) and the rejection
number (ri), until a decision is made. Sometimes
acceptance is not allowed at the early stages of
multiple sampling. However, the lot's rejection can
occur at any stage of the multiple sampling.
Finally, there are some specific sampling plans in
use in quality inspection, such as the Sequential
sampling plan and the Skip lot sampling plan. The
Sequential sampling plan represents the ultimate
extension of multiple sampling. Only a fraction of
the submitted lots is inspected by using the Skip lot
sampling plan.

Testing the 1st sample

(n1 = 45 )

x c1
(x = 0)

x > c2
(x > 3)

If the number
of defectives
(x)
c1 < x c2
(0 < x 3)
Testing the 2nd sample
(n2 = 2n1 = 90 )

x1 c 2
(x1 3)

If sum
of defectives (x1)
in both samples
(n1 + n2)

Accept the lot

3 An efficient measure for a multiple

sampling scheme

x1 > c2
(x1 > 3)

Considering economically, it is important to deal

with a rational sample size when inspecting a lot,
because a huge number of inspected items would
result in high costs.
If single sampling is carried out, only one sample
is selected from a lot. The number of items needed
in the case of multiple sampling scheme may vary

Reject the lot

End

Fig. 1 Sampling scheme for double Philips sampling plan

E-ISSN: 2224-2678

722

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

1
0,9
0,8
0,7
0,6
P a 0,5
0,4
0,3
0,2
0,1
0

from trial to trial. The Average Sample Number

(ASN) represents an average of what might happen
over many trials with a fixed incoming defect level
(p). ASN measures the efficiency of a multiple
sampling scheme. For a single LASP (n,c) each lot
(Ni) is known and has a known sample of size (ni)
taken and inspected. However, for double, multiple
and sequential LASPs the amount of sampling
varies depending on the number of defectives
observed. A long term ASN can be calculated for
any given double, multiple or sequential plan
assuming that all lots have a defect level of (p).
Using the ASN curve, a plot of the ASN versus the
incoming defect level (p) describes the sampling
efficiency of a given LASP scheme. Since the
number of samples in the case of multiple sampling
may vary from trial to trial, the ASN represents the
average of what might happen over many trials with
a fixed level of defectives in the incoming lots. A
formula for the ASN of a double sampling plan is:
ASN = n1 P1 + (n1 + n2 )(1 P1 )

0.01 0.02 0.03 0.04 0,05 0,06 0,07 0,08 0,09 0,10

Percent of defectives (P d )

Fig. 2 The OC curve (for the LASP: n = 52, c = 3)

The probability of observing exactly

defectives is given by the following formula:

f (d ) = P (d ) = p d (1 p )

nd

(1)

(2)

Obviously, this formula corresponds to the

binomial distribution. Furthermore, the probability
of a lot acceptance is the probability that the number
of defectives (x) is less than or equal to the
acceptance number (c). This means that:

where P1 is the probability of a decision on the first

sample.

4 The properties of the Operating

Characteristic (OC) Curve

Pa = p x (1 p )

n x

x=0

The operating characteristic (OC) curve is an

important tool for explaining and researching the
properties of LASPs. The OC curve plots the
percent of defectives on the x-axis versus the
probability of accepting the lot on the y-axis. In fact,
the OC curve shows the probability of acceptance
depending on the percent of defectives, with the
precondition that each lot contains defectives. In an
ideal case, after using a chosen LASP, all good lots
will be accepted and all bad lots will be rejected. On
the other hand, since the decision on acceptation or
rejection of the lot depends on the sample taken
from the lot, there is a real possibility of making the
wrong choice in inspection praxis.
An example of the OC curve is shown in Fig. 2.
Regarding the OC curve the following is valid:
the lot size (N) is very large compared to the
sample size (n), therefore, removing the sample
doesnt significantly change the remainder of the
lot (no matter how many defectives are in the
sample which is taken at random),
the distribution of the number of defectives (d) in
a random sample of (n) items is approximately
binomial with parameters (n) and (p).

E-ISSN: 2224-2678

n!
d!(n d )!

(d)

n!
x!(n x)!

(3)

where Pa = P {x c} .
The following combinations of values for (pa, pd)
are calculated using Pd = 0.01,, 0.12, n = 52 and c
= 3 in (2) and (3):
Pa

Pd

Pa

Pd

Pa

Pd

0.998
0.980
0.930
0.845

0.01
0.02
0.03
0.04

0.739
0.620
0.502
0.394

0.05
0.06
0.07
0.08

0.300
0.223
0.162
0.115

0.09
0.10
0.11
0.12

These data represent the coordinates of the points at

the OC curve for a sampling plan n = 52, c = 3.
The calculation of a sampling plan with a given
OC curve is based on complex formulas. Let us
design such a sampling plan that the probability of
acceptance is 1- for the lots with fraction defective
(p1) and the probability of acceptance is () for the
lots with fraction defective (p2). Typical notions for
such a case are:
is the suppliers risk that a good lot is rejected
(type I error),
is the consumers risk that a bad lot is accepted
(type II error),

723

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

p1 is the Acceptable Quality Level (AQL),

p2 is the Lot Tolerance Percent Defective
(LTPD).
If binomial sampling is valid, the possibility of
lot acceptance with defectives (p1), for the sample
size (n) and the acceptance number (c), is:
d

nd

d =0

n!
d ! (n d )!

1.0

(4)

pl
es
e
iz
es

pl

am

e
iz

AOQL

(5)

Consumers risk

pa

where Pa = .
Obviously, there is no direct (simple) calculation
solution because (4) and (5) are nonlinear
simultaneous
equations.
However,
iterative
techniques are available that give good approximate
solutions.
The lot's inspection in praxis is often done by the
consumer's inspection department after the lots were
received from the supplier. Assuredly, the lots will
not always contain the same percent of defectives,
and it is necessary to use the OC curve regardless of
the fact that the method does not completely ensure
that the accepted lot will be good, and there is also a
certain possibility that a good lot can be rejected.
The OC curve is just a picture of a certain
sampling plan, so that each sampling plan has a
unique OC curve. The OC curve is defined by the
sample size and the acceptance number. The shapes
of ideal and real (round) OC curves are shown in
Fig. 3.
In praxis, it is important to know how the sample
responds to a range of percentages of defectives in
the lot. The OC curve includes this basic
characteristic of each sampling plan. Nevertheless, it
should be accentuated that this method does not
guarantee with 100% that each accepted lot is a
good lot. Moreover, there is a certain probability
that a good lot is denied. In fact, the probability of
occurrence of certain defectives in the lot can be
obtained using the OC curve.
It is necessary to explain the values shown in
Fig. 3. Acceptable Quality Level AQL (pa) is the
maximal percent of nonconforming items, or the
maximal number of nonconformities per hundred
items, which is considered a satisfying process
average for inspection purposes. Therefore, it is not
lot specific. The AQL is the maximal percent of

E-ISSN: 2224-2678

Real (round) OC curve

am

n!
d ! (n d )!

s
er
rg

d =0

nd

la

Pa = p2 (1 p2 )
d

rs
le

Ideal OC curve

al

where Pa = 1 .
Furthermore, the possibility of lot acceptance
with defectives (p2) is:

Suppliers risk

sm

ProbabilityY-Axis
of acceptance

Pa = p1 (1 p1 )

defectives that is acceptable for both producer and

consumer, i.e. the poorest level of quality (percent
of nonconformities) that the process can tolerate.

pn

pt

Fraction defective (%)

AQL

Fig. 3 The shape of OC curve

The producer prefers to design a sampling plan

so that there is a high probability of accepting a lot
that has a defect level less than or equal to the AQL.
The AQL is stated in the standard as a percent (so,
an AQL of 0.2 is a rate of 0.2 nonconforming items
per 100 items or 0.2%). For example, if the AQL is
2.0 then the resulting product is acceptable to both
parties, i.e. the producer is ready to produce
approximately 2% of defectives. Furthermore, the
Neutral Quality (pn) is an indifference quality level:
i.e. there is a 50% chance that the bad lot can be
accepted and a 50% chance that the good lot can be
rejected. Finally, the Average Outgoing Quality
Level (AOQL) represents the maximal possible
rectangle under the real OC curve. The AOQL is the
worst possible quality that results from the
rectifying inspection programme. It is the maximum
AOQ for all possible levels of incoming quality.
Furthermore, the next rule is valid: the round OC
curve is for smaller values (n) and (c) while the
ideal OC curve is for bigger values (n) and (c).
When the round OC curve is used, there is a certain
probability that the entire lot is rejected if it contains
less defectives than LTPD, however it is also
possible that the lot is accepted if it contains more
defectives than LTPD.
With the OC curve there is a probability that the
lot is rejected if the defect level is less than the
AQL, but also that the lot is accepted if the defect
level is higher. There is a possibility that the lot
contains less defectives than the AQL, but -risk
(the suppliers risk) still exists. Such a case occurs at

724

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

the acceptable quality level (pa). Usually, -risk is

5%. This is a type I error (), which is the
probability for a given (n,c) sampling plan of
rejecting a lot that has a defect level equal to the
AQL.
On the other hand, there is a possibility that a lot
with an unacceptable quality is accepted. Obviously,
the consumer suffers when such a risk occurs. This
risk occurs at the tolerance quality level (pt).
Usually, -risk is 10%. This is a type II error (),
which is the probability for a given (n,c) sampling
plan of accepting a lot with a defect level equal to
the LTPD.
Regarding decision on a lot, four possible
outcomes are shown in Table 1.
Generally, the OC curve is characterized by its
AQL and LTPD. In fact, the AQL describes what
the sampling plan generally accepts. Formally, it is
the percent of defectives with a 95% chance of
acceptance. The LTPD describes what the sampling
plan generally rejects. Formally, it is the percent of
defectives with a 10% chance of acceptance. The
LTPD is a designated high defect level that would
be unacceptable to the consumer. By using the
LASP and AQL, the producer is ready to produce
the number of defectives that will be accepted by
the consumer.

than 5% of LTPD in order to achieve a reasonably

small probability of rejecting the good lot if c = 0 for
the chosen LASP. In the case that this average
cannot be ensured one should choose another type
of LASP with a bigger sample size, because the
LASP which has c = 0 is obviously not a good
choice.
As mentioned beforehand, the LASP is designed
in such a way that the OC curve passes through two
designated points corresponding to AQL and LTPD.
This presents a problem if an error occurs.
Nevertheless, some corrections of the acceptance
procedure can be done in order to ensure that the
OC curve passes through the aforementioned points.
This can be shown, but only if the OC curve passes
through (AQL, 1-) and (LTPD, ). The LASP's
errors can have a huge influence on the sample size
needed to achieve the desired position of the points
on the OC curve [19]. The sampling plans can be
used for a variety of purposes depending on past
history and other circumstances. One must provide
justification for the AQLs and LTPDs used in the
quality control department. This requires that the
aim of each inspection is defined.
For example, an AQL of 1.0% is specified as the
aim of inspection of the main defects. This AQL is
not necessarily equal to the sampling plan AQL (so
it is noted as 'AQL-Alternative' to make this
distinction clear). The AQL-Alternative is
interpreted as the maximum percent defective for
which acceptance is desired. However, the AQLAlternative does not mean that there is a
permission to deliver bad lots.
All lots above the AQL-Alternative are best
rejected. On the other hand, all lots below the AQLAlternative are best accepted. The break-even
quality between acceptance and rejection is,
therefore, represented by the AQL-Alternative. The
cost of carrying out a 100% inspection will exceed
the benefits of doing so in terms of fewer defects
released if lots come with percent defectives below
the AQL-Alternative. All lots should be 100%
inspected if a manufacturing process always
produces lots with percent defectives above the
AQL-Alternative. The quality control department
could use a sampling plan to screen out lots not
requiring 100% inspection if some lots are below
the AQL-Alternative.
A sampling plan with a LTPD the same as the
AQL-Alternative can be used to ensure that lots
worse than the AQL-Alternative are rejected, but
with the risk of rejecting some acceptable lots. The
single sampling plan with n = 230 and c = 0 (which
has a LTPD of 1.0%) is appropriate for an AQLAlternative of 1.0%.

Table 1 Outcomes in inspection praxis

Suppliers risk of
rejecting good lot
( risk; type I error).
Consumer's risk of
accepting bad lot
( risk; type II error).
Consumer's
(purchaser's)
decision

Supplier's (producer's)
activity
The lot
conforms

The lot does

not conform

Accept

-risk

Reject

-risk

Normally, each consumer prefers a sampling

plan with a low probability of accepting a lot with a
defect level the same as the LTPD. The LTPD is an
important criterion for the LASP [17], which can be
used in the case of limited resources in the quality
control department when a minimal size of samples
is used.
The LASP is based on a clearly defined
correlation with the lot size. Generally, a lot is
rejected if any defectives are found in a sample. The
procedure gives the proportion of the lot which must
be sampled. The Schilling Table [18] can be used
for this procedure. The manufacturing process
should be run at the average quality level of less

E-ISSN: 2224-2678

725

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

Three cases are possible:

all lots are inspected and the amount to be
inspected is (N) if all inspected items are
defective,
no lot is rejected if all inspected samples (ni)
from the lots (Ni) contain zero defectives,
if the lot quality is 0 < p < 1, the average number
of the inspected items per lot vary between the
sample size (n) and the lot size (N).
The Average Total Inspection (ATI) denotes the
average number of units that will be inspected for a
particular incoming quality level and probability of
acceptance. ATI presents the expected number of
units inspected after rectifying inspection for a
given quality level.
If all rejected lots are 100% inspected, it is easy
to calculate ATI when lots come consistently with a
defect level of (p). For a LASP (n,c) the formula for
ATI is:

For example, the required sample size for zero

acceptance number and desired confidence level of
95% is: n = 300 / AQL-Alternative. Finally, the
following rule is valid: the AQL should be used
when producing many lots of a product, and the
LTPD should be used when producing smaller
number of lots.
If sampling and testing is non-destructive, a
common procedure is to use the 100% inspection of
the rejected lots and to replace all defectives with
good units. In such a case, all rejected lots no longer
contain any defectives and the only defectives left
are those in the lots that were accepted during the
sampling procedure as good ones.
Assume that all lots come in with exactly a
proportion of defectives (p). After having inspected
a rejected lot, the final fraction defective is zero for
that lot. Nevertheless, the accepted lots have a
fraction defective (p). Therefore, the outgoing lots
from the quality control department are a mixture of
the lots with a fraction defective (p) and zero.
Each LASP ensures a certain average quality as a
result of all received lots. The Average Outgoing
Quality (AOQ) is the expected average quality of
outgoing items for a given value of incoming
product quality. The AOQ can be expressed as
(process average)(probability of acceptance)106,
where 106 is a conversion to parts per million (ppm).
The AOQs refer to the long term defect level for
the combined LASP and 100% inspection of
rejected lots process.
If all delivered lots have a defect level of exactly
(P), and the chosen LASP (n,c) indicates a
probability (Pa) of accepting such a lot, over the
long run the AOQ can be calculated as:
AOQ =

[Pa P (N n )]

ATI = n + (1 Pa )(N n )

where Pa is the probability of accepting a lot, N

is the lot size and n is the sample size.
After replacing the defectives, and if during the
replacement procedure some type of error occurs,
the formula for ATI is:
ATI =

1
[n + (1 Pa ( e ) ) (N n )]
1 pe

(8)

If defectives are not replaced, the formula for

ATI is:
ATI = n + (1 Pa ( e ) ) (N n )

(9)

Formula for the ATI curve for double sampling

plans is:

(6)

ATI = Pa1n1 + Pa2 (n1 + n2 ) + N (1 Pa )

If the sample size is assumed to be very small

compared to the lot size, the AOQ is practically the
same as the product of Pa and P.
For example, let N = 10,000, sampling plan with
n = 52 and c = 3. Let the quality of incoming lots be
0.03. It follows that Pa = 0.930 for P = 0.03, based
on the table of the OC curve. Furthermore, by using
(6) the calculated AOQ is 0.0278.

(10)

where Pai is the probability of acceptance on the

ith sample.
The calculation of some values of ATI is
presented in Table 2.
Table 2 The calculation of the ATI using some different
values of P
P
0.05
0.04
0.03
0.02
0.01

5 The average amount of inspection

per lot
An important question of acceptance sampling is:
what is the total amount of inspection when rejected
lots are screened?

E-ISSN: 2224-2678

(7)

726

52

10,000

Pa
0.739
0.845
0.930
0.980
0.998

1 - Pa
0.261
0.155
0.070
0.020
0.002

N-n

9,948

ATI
2,648
1,594
748
251
72

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

curves present a special case when c = 0), are

defined by the OC curves and they have practically
the same (LTPD, ) point as the corresponding
ANSI/ASQ Z1.4 single normal plan. The c=0 plans
set = 0.1. It tries to match the ANSI/ASQ Z1.4
plans at the RQL (or LTPD) point. However, the
main difference is that the c=0 plans are oriented to
the LTPD whilst ANSI/ASQ Z1.4 are oriented to
the AQL which is set by a quality engineer in praxis
(the suppliers process average should be less than
the chosen AQL). Furthermore, the calculations in
the ANSI/ASQ Z1.4 use the binomial or Poisson
distribution, while they use the hyper-geometric
distribution in the c=0 plan. Considering the OC
curves for both plans, a clear advantage of using
much smaller sample sizes in the case of c=0,
compared to the corresponding ANSI/ASQ Z1.4
plans, is evident. For example, if N = 1,300 and
AQL = 4.0%, the sample size (n) is 125 in the case
of ANSI/ASQ Z1.4 but only 18 in the case of c=0.
Consequently, c = 10 and c = 0. Hence, the
inspection costs drop from 125 to 18 items if the
c=0 is used instead of the corresponding plan
ANSI/ASQ Z1.4.
It should be mentioned that ANSI/ASQ Z1.4 is
probably the most commonly used standard for the
attribute acceptance case in the last ten years. This
standard offers seven levels of inspection: the
reduced, normal and tightened inspections as well as
four levels of special inspection (that should be used
only when small sample sizes are necessary and
large risks can be tolerated).
Selecting a plan from ANSI/ASQ Z1.4-2003
implicates the step-by-step procedure: select an
AQL; select the inspection level; determine the lot
size (n); find the appropriate sample size code letter
in the table of the standard; determine the type of
sampling plan single, double or multiple; enter the
appropriate table to find the plan to be used by using
the selected AQL and sample size code letter;
determine the normal, reduced and tightened plans
as required from the corresponding tables. A more
detailed explanation of the sampling plan procedure
is outside the scope of this paper.
A comparison of some double sampling plans is
provided in Table 3.
By observing data shown in Table 3, it can be
concluded that MIL-STD 105E protects the
supplier, Dodge-Romig protects the consumer,
while Philips is somewhere in between.
Considering the (n1 + n2) criterion, Philips is the
most demanding plan with n1 + n2 = 405. DodgeRomig follows with n1 + n2 = 370 for pt and 385 for
AOQL. MIL-STD 105E with n1 + n2 = 250 is the least
demanding plan.

Example:
What is the value of ATI for the chosen sampling
plan n = 52 and c = 3 if N = 10,000 and P = 0.04?
By using the coordinates for the points at the
specified sampling plan OC curve, it can be noticed
that P = 0.04 matches Pa = 0.845. Hence, it follows
that ATI = 1,594 by using (7). It means that a total
of 1,594 items was inspected for the considered
sampling plan, i.e. as high as 15.94% of the whole
lot (N). Obviously, this would result in unacceptably
high costs. Therefore, it is much better to use a
smaller P: for example, ATI = 251 if P = 0.02 or
even better ATI = 72 if P = 0.01.

6 Parameters' comparison for

different LASPs
Sampling by attributes is a method of quality
inspection in which every 100 items is classified by
percentage of conformance or non-conformance.
For example, 3 non-conforming would be a 3% rate.
Acceptance sampling is used to determine whether
to accept or reject a specific quantity of goods or
materials [20]. Philips, Dodge-Romig and MIL-STD
105E form a group of classic sampling plans used in
the sampling inspection praxis over many years. As
a matter of fact, MIL-STD 105E was cancelled three
times: first in 1995, then in 2001 and finally in
2008. MIL-STD 105E was simply replaced by
ANSI/ASQC Z1.4 in such a way that Z1.4 was
actually a reprint of 105E. MIL-STD 105E was
cancelled to reduce costs through the elimination of
duplication. Thereafter, ANSI/ASQ Z1.4-2008 with
its sampling procedures and tables for inspection by
attributes replace ANSI/ASQ Z1.4-2003 in 2008.
The ANSI/ASQ Z1.4-2008 follows its previous
(2003) version, so the procedures and tables remain
practically unchanged from MIL-STD 105E.
Moreover, other standards such as ISO2859-1,
DIN40080, NF06-022 and BS6001 exist as
equivalents of the ANSI/ASQ Z1.4.
In fact, MIL-STD-105E is nearly the same as
ANSI/ASQ Z1.4 (1993) in relation to the tables.
ANSI/ASQ Z1.4 has changes in the explanatory text
that do not affect procedures, however, it has
different switching rules. ANSI/ASQ Z1.4 uses the
limit numbers for switching and it contains some
additional OC curves that describe the protection
provided by the switching procedure for the periods
of constant quality.
The so-called 'c=0' described in [21] by Squeglia
presents broadly used sampling plans in the attribute
inspection today. The c=0 plans, in which the
acceptance number is zero in all cases (so all OC

E-ISSN: 2224-2678

727

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

Table 3 Parameters of the double Philips, Dodge-Romig

and MIL-STD 105E sampling plans
LASP
type

saves money and time without influencing the AOQ

of the lots.
Showing how the right chosen LASP can ensure
a high reliability of acceptance of a large number of
quality items is a contribution of this research.
Finally, it should be pointed out that it is of
significant importance to know the LASPs'
characteristics as well as their comprehensive
mathematical base, especially due to the existing
differences between the considered sampling plans.

DOUBLE SAMPLING PLAN

PHILIPS

DODGE-ROMIG

Lot size

MIL-STD
105E

N = 3,500

Criterion

pn (%)

pt (%)

AOQL (%)

pa (%)

1st sample

n1=135
c1=0

n1=140
c1=0

n1=145
c1=0

n1=125
c1=0

2nd sample

n2=270
c2=3

n2=230
c2=3

n2=240
c2=3

n2=125
c2=3

AOQL (%)

0.5

0.52

0.5

0.62

pa (%)

0.35

0.4

pn (%)

pt (%)

1.9

References:
[1] J.O. Westgard, P.L. Barry, M.R. Hunt, T.A.
Groth, Multi-Rule Shewhart Chart for Quality
Control in Clinical Chemistry; Clin Chem, vol.
27, No. 3, 1981, pp. 493-501.
[2] C.H. deVerdier, T. Groth, J.O. Westgard, What
is the Quality of Quality Control Procedures,
Scand J Clin Lab Invest, vol. 41, 1981, pp.1-14.
[3] C.H. Chen, Economic Design of Dodge-Romig
AOQL Single Sampling Plans by Variables
With the Quadratic Loss Function, Tamkang
Journal of Science and Engineering, vol. 8, No.
4, 2005, pp. 313-318.
[4] C.H. Chen, C.Y. Chou, Economic Design of
Dodge-Romig Lot Tolerance per Cent
Defective Single Sampling Plans for Variables
Under Taguchis Quality Loss Function, Total
Quality Management, vol. 12, No. 1, 2001, pp.
5-11.
[5] H.F. Dodge, H.G. Romig, Sampling Inspection
Tables, John Wiley, New York, NY, 1959.
[6] H.C. Hamaker, Some Basic Principles of
Sampling Inspection by Attributes, Applied
Statistics, vol. 7, No. 3, 1958, pp. 149-159.
[7] R.A. Banzhoof, R.M. Brugger, Reviews of
Standards and Specifications MIL-STD-1235
(ORD), Single and Multi-level Continuous
Sampling Procedures and Tables for Inspection
by Attributes, Quality Technology Journal, vol.
2, No. 1, 1970, pp. 41-53.
[8] J.M. Juran, A.B. Godfrey, Juran's Quality
Handbook, 5th ed., McGraw-Hill, New York,
NY, 1999.
[9] E. Banovac, I. tritof, Analysis of Applicable
Methods of Incentive Regulation in the Field of
Distribution
of
Electricity,
WSEAS
Transactions on Systems, vol. 4, issue 8, 2005,
pp. 1224-1232.
[10] E. Banovac, . Bogdan, I. Kuzle, Choosing the
Optimal Approach to Define the Methodology
of a Tariff System for Thermal Energy
Activities, Strojarstvo, vol. 49, No. 6, 2007, pp.
409-420.

Furthermore, AOQL is 0.62% for MIL-STD

105E, while for the other two plans AOQL is almost
the same (0.5 or 0.52%). It should be emphasized
that ANSI/ASQ Z1.4 is the same as MIL-STD 105E
by the consumer's protection criterion in this
consideration due to the fact that Z1.4 is just a
mirror version of 105E as mentioned before.
Furthermore, the c=0 sampling plans provide equal
or even greater consumer protection with less
overall inspection compared to the corresponding
MIL-STD-105-E.
At the end of this section, it should be stressed
that there is a need to consider the total cost of using
the sampling plan which includes the costs of
inspection, reworks and defective items returned by
the consumers [22].
A broader context related to the topic considered
in this paper is presented in the following
noteworthy papers [23]-[31].

7 Conclusion
The acceptance sampling is very useful when
inspecting large lots. As shown in this paper,
different Lot Acceptance Sampling Plans (LASPs)
developed for the attribute acceptance sampling
represent a powerful inspection tool used by the
quality control experts.
In praxis, it is often too expensive and even
practically impossible to use 100% inspection when
huge purchased lots should be inspected by the
quality control staff. Hence, the authors advocate for
the LASPs' use when large lots are inspected in
industrial branches, such as the electric-power or
gas branch as well as in medical laboratory
procurement. In fact, the ordinary usage of LASPs

E-ISSN: 2224-2678

728

Volume 13, 2014

WSEAS TRANSACTIONS on SYSTEMS

Sejid Tenjak, Ivan Banovac

[23] I. Dafi, M. Glavi, S. Tenjak, A

Component-Based Power System ModelDriven Architecture, IEEE Transactions on
Power Systems, vol. 19, issue 4, 2004, pp.
2109-2110.
[24] I. Dzananovic, S. Tenjak, On Some Technical
Aspects of Transmission and Distribution
Networks Interaction, Proc. of the 7th IASTED
International Conference on Power and Energy
Systems, Palma de Mallorca, Spain, 2007.
[25] J. Hynek, V. Janeek, L. Svobodov, Problems
Associated with Investment in Advanced
Manufacturing
Technology
from
the
Management Point of View, WSEAS
Transactions on Systems, vol. 8, issue 6, 2009,
pp. 753-762.
[26] M. Camilleri, F. Neri, M. Papoutsidakis, An
Algorithmic Approach to Parameter Selection
in Machine Learning using Meta-Optimization
Techniques, WSEAS Transactions on Systems,
vol. 13, 2014, pp. 203-212.
[27] M. Papoutsidakis, D. Piromalis, F. Neri, M.
Camilleri, Intelligent Algorithms Based on
Data Processing for Modular Robotic Vehicles
Control, WSEAS Transactions on Systems, vol.
13, 2014, pp. 242-251.
[28] M. Zahran, Smart Grid Technology, Vision,
Management
and
Control,
WSEAS
Transactions on Systems, vol. 12, issue 1, 2013,
pp. 11-21.
[29] O. Kuljaa, S. Tenjak, V. Koroman,
Performance Analysis of Adaptive Neural
Network Frequency Controller for Thermal
Power Systems, Proc.of the 9th WSEAS
International Conference on Automatic
Control, Modeling & Simulation, Istanbul,
Turkey, 2007.
[30] S.A. Staines, F. Neri, A Matrix Transition
Oriented Net for Modeling Distributed
Complex Computer and Communication
Systems, WSEAS Transactions on Systems, vol.
13, 2014, pp. 12-22.
[31] V. Vlad, C. Buzduga, C. Ciufudean, An
Approach to Developing Power Grid Control
Systems with IEC 61850, IEC 61499 and
Holonic Control, WSEAS Transactions on
Systems, vol. 13, 2014, pp. 503-509.

[11] E. Banovac, T. Gelo, J. imurina, Analysis of

Economic Characteristics of a Tariff System
for Thermal Energy Activities, Energy Policy,
vol. 35, issue 11, 2007, pp. 5591-5600.
[12] J. Shade, Acceptance Sampling in Quality
Control, J. Roy. Stat. Soc. A Sta., vol. 174, No.
4, 2011, pp. 1185-1186.
[13] K. Dumicic, V. Bahovec, N.K. Zivadinovic,
Studying an OC Curve of an Acceptance
Sampling: A Statistical Quality Control Tool,
Proc. of the 7th WSEAS Intern. Conf. on
Mathematics & Computers in Business &
Economics, Cavtat, Croatia, 2006, pp. 1-6.
[14] H.M. Judi, R. Jenal, C.L. Chun, Product
Inspection Using Computer-aided Acceptance
Sampling for SMEs, Eur. J. Sci. Research, vol.
27, No. 1, 2009, pp. 6-15.
[15] H.C. Hamaker, R.V. Strik, The Efficiency of
Double Sampling for Attributes, American
Statistical Association Journal, vol. 50, No.
271, 1955, pp. 830-849.
[16] E. Banovac, D. Kozak, An Analytic Review of
the Characteristics of the Lot Acceptance
Sampling Plans Used for Acceptance of Large
Lots, International Review of Electrical
Engineering (I.R.E.E.), vol. 3, No. 6, 2008, pp.
1070-1076.
[17] E.G. Schilling, A lot Sensitive Sampling Plan
for Compliance Testing and Acceptance
Inspection, Quality Technology Journal, vol.
10, No. 2, 1978, pp. 47-51.
[18] E.G. Schilling, Acceptance Sampling in Quality
Control, Marcel Dekker, New York, NY, 1982.
[19] E. Banovac, I. Kuzle, Applicability of the
LASPs in the Electric-Power Industry, Proc. of
the International IEEE Conference EUROCON
'09, Saint Petersburg, Russia, May 18-23, 2009,
pp. 1152-1157.
[20] L.J. Krajewski, L.P. Ritzman, Operations
Management, Processes and Value Chains, 7th
ed., Prentice Hall, NJ, 2005.
[21] N.L. Squeglia, Zero Acceptance Number
Sampling Plans, 5th ed., ASQ Quality Press,
Milwaukee, Wisconsin, 2008.
[22] A. Haji, R. Haji, The Optimal Policy for a
Sampling Plan in Continuous Production in
Terms of the Clearance Number, Computers
and Industrial Engineering Archive, vol. 47,
No. 2-3, 2004, pp. 141-147.

E-ISSN: 2224-2678

729

Volume 13, 2014